A note on Kim's Korean question particles seen as pronouns

Michal Starke
2011-03 [minor edits 2012-09]
To mark some clauses as interrogative, Korean appends a special marker to the end of the clause. In those cases, Korean interrogatives differ from declaratives only in the shape of their final markers:
ni-ka ppang-ul mek-ess ta

you-nom bread-acc eat-past decl

“You ate the bread”
(2) a. ni-ka ppang-ul mek-ess ni

you-nom bread-acc eat-past Q

b. ni-ka ppang-ul mek-ess nya

you-nom bread-acc eat-past Q

“Did you eat the bread?”

c. ni-ka ppang-ul mek-ess na

you-nom bread-acc eat-past Q

d. ni-ka ppang-ul mek-ess ka

you-nom bread-acc eat-past Q

roughly: “Did you eat the bread, I wonder...”

Kim (2011) observes that the four question particles in (2) are identical to pronouns:  they look like pronouns and their semantic interpretation corresponds to the interpretation of their pronominal counterpart (see below). Accordingly, Kim makes the elegant proposal that Korean question particles are in fact pronouns which express the addressee of the question. The beauty of this proposal however comes at a serious cost:

The point of this small note is to show that we can keep the elegance of Kim's proposal and solve both problems, once we reason in terms of phrasal spellout with its superset effect (Starke 2002, 2009, Caha 2009).

The coarse semantic properties of the particles.

The four questions in (2) have different semantics, along roughly two dimensions. First, some of the questions are full-blown questions addressed to the interlocutor, whereas others express questions that are not meant to be answered by the interlocutor - somewhat akin to "I wonder..." contexts. Second, those particles that address the interlocutor necessarily come with a particular register since all interlocutor-addressed speech in Korean is coded for the relative social status of the interlocutors. The four particles above are characterised as follows:
(3) a.
ni +hearer-oriented low level

nya +hearer-oriented low level

na -hearer-oriented

ka -hearer-oriented

Kim's major observation is that those morphemes correspond to independently existing Korean pronouns, with matching properties:

(4) a.
ni 2nd person pronoun low level

na 1st person pronoun

ka 3rd person pronoun

The missing nya particle corresponds to the composition of ni and the vocative particle ya: ni+ya > nya. As is obvious from (3-4), the question particles correspond to pronouns both phonologically and semantically: exactly those questions that are hearer-oriented correspond to hearer-oriented pronouns, ie 2nd persons, also matching their speech level. Kim also documents that nya does indeed inherit the semantics of vocative, due to the presence of the vocative morpheme ya, and that na and ka differ in the predicted direction given their differing pronominal roots.

So why do question particle look so much like pronouns? This is what is expected under the performative hypothesis (Ross 1970), according to which every question is underlyingly of the form “I ask ADRESSEE whether p”, if the question particles are in fact pronouns identifying the addressee. Kim's solution is thus that (i) the performative hypothesis is syntactically correct, and questions have not only an interrogative feature, but also an adressee slot, and (ii) the so-called question-particles in (3) fill that slot. This is a very interesting new view on question particles.

The dilemma of identifying particles with pronouns and its solutions.

One problem with this solution is that it does not actually capture the connection between pronouns and question particles. What it captures is a wider - and inaccurate - generalisation: pronouns will occur as sentence-final particles in any clause-type containing an addressee, interrogative or not. This is because sentences other than questions also have an addressee slot and Kim's approach predicts pronouns to fill those slots too. Both in Ross' (1970) performative approach and in its modern adaptations such as Truckenbrodt (2006), declaratives also have an addressee slot (roughly, “I tell ADRESSEE that p”), and hence if pronouns simply fill the addressee slot, they are predicted to occur in declaratives as well as interrogatives. Which is not the case. (Mutatis mutandis for imperatives, etc.)
The problem cannot be solved in the obvious way: we could try to encode the fact that pronouns occur as sentence-final particles only in interrogative clauses by adding (the counterpart of) an interrogative feature to them, but in doing so we would lose the identity between a pronoun (no interrogative feature) and a sentence-final particle homophonous to a pronous (with an interrogative feature). So at first sight, we can't have it both ways.
But we can. What we want to express is that there is a single lexical entry covering both pronouns and question markers, and that questions markers are a superset of pronouns, ie pronouns + interrogative features. The superset effect in nanosyntax (Starke 2002, 2009, Caha 2009) yields just that. The superset effect derives from the fact that a syntactic structure can be spelled out by a lexical item if the syntactic structure is contained in the lexical entry for that item (and hence if the lexical item is a superset of the syntactic material). Thus if the syntactic lexical entry for the question particles is the syntactic equivalent of [Q [ pronoun]], it will match both syntactic contexts: qua question particle it will spell out [Q [ pronoun]] (and not just [pronoun], contra Kim); qua pronoun, it will spout out the syntactic constituent [pronoun]. This resolves the dilemma: there is a single lexical entry, thus capturing the fact that the question particle and the pronoun have the same phonological shape, but that lexical entry matches two distinct syntactic structures, capturing the fact that the entry is restricted to typing questions. 1

Mismatches between pronouns and question particles.

Apart from resolving the above dilemma, this solution has the advantage of explaining cases where the match between the question particle and the pronoun is not perfect. Such cases are attested: for instance in some dialects of Korean the question particle 'ka' does not correspond to an existing productive pronoun. In some of these dialects the third person pronoun alternates between 'ka' and 'kyay', in others 'ka' has become archaic/formal and 'kyay' is standard. But in none of these dialects does the question particle become 'kyay', it always remains 'ka'. Hence within those grammars, the question particle cannequated with the pronoun.

Phrasal spellout and its 'superset' effect predict just that: in these dialects, the lexical entry for 'ka' is the same as before, namely [Q [pronoun]]; but contrary to before, this entry now competes with another lexical entry" [pronoun] for 'kyay' (again skipping the internal structure of the pronoun, irrelevant for the present purposes). In these grammars, a question particle addressed to the hearer works as before: 'ka' is the only candidate to spell it out since it is the only lexical item to contain Q. When syntax builds a second person pronoun, things comes out differently: because of the superset effect, there are now two candidates to spellout [pronoun], 'ka' and 'kyay'. At this point, the elsewhere principle comes into play and the more specific 'kyay'=<+pronoun> wins the insertion race (see Starke 2002, 2009, Caha 2009 for details). Hence in these dialects, a pronoun will come out as 'kyay' and not as 'ka', despite the lexical entry of 'ka' being [Q [pronoun]].

These mismatches thus come out for free, in fact they are predicted by the approach: when a language has a lexical item for [Q [pronoun]], it can chose to add another entry for [pronoun] -- or not. The general rules of spellout will do the rest. Our rephrasing of the Kim's analysis in terms of phrasal spellout thus allows us to preserve the elegance of his proposal -- interrogative particles are none others than the lexical entry for pronouns -- and at the same time capture the fact that these pronouns are restricted to interrogative clause-typing and can drift away into separate lexical entries.


1. The same effect might be achieved via the subset principle of Distributed Morphology, depending on the lexical entries for the other 50 or so Korean particles. The lexical entry for the particle would be the pronoun, without any interrogative feature, and it would have to be the best match for the combination of Q and pronoun, but not the best match for declarative+pronoun, etc. All the other particles would thus have to be more specific and block the pronouns. It is unclear whether this is plausible, see for instance Kim's discussion of the clausal particle kka. No such difficulty arises under the superset formulation.


Caha  (2009) The Nanosyntax of Case, PhD Thesis, University of Tromso, http://ling.auf.net/lingBuzz/000956.
Kim, Chonghyuck (2011) Korean question particles are pronominals: A transparent case of representing discourse participants in the syntax, http://ling.auf.net/lingBuzz/001157
Ross, John R. (1970) On Declarative Sentences. In Readings in English Transformational Grammar, eds. Roderick A. Jacobs and Peter S. Rosenbaum, 222-272. Ginn, Walthan: MA.
Starke, Michal (2002) The day syntax ate morphology. Class taught at the EGG summerschool, Novi Sad.
Starke, Michal (2009) Nanosyntax. A short primer to a new approach to language. Nordlyd 36.1, ed. Peter Svenonius, Gillian Ramchand, Michal Starke, and Knut Tarald Taraldsen, pp. 1– 6. CASTL, Tromsø.